← Về thư mục
verl/utils/dataset/rm_dataset.py
- download_files_distributed · function · L27-L37 — def download_files_distributed(download_fn)
- RMDataset · class · L40-L143 — class RMDataset(Dataset)
- init · method · L42-L68 — def init(self, parquet_files: Union[str, List[str]], tokenizer, prompt_key='prompt', chosen_key='chosen', rejected_key='rejected', max_length=1024, add_eos=True, cache_dir='~/.cache/verl/rm')
- _download · method · L70-L83 — def _download(self)
- _download_files · function · L72-L81 — def _download_files()
- _read_files_and_tokenize · method · L85-L94 — def _read_files_and_tokenize(self)
- len · method · L96-L97 — def len(self)
- _pad_to_length · method · L99-L112 — def _pad_to_length(self, input_ids, attention_mask)
- getitem · method · L114-L143 — def getitem(self, item)